#45Live ↗AI Agents & Automations

PipelineRiver.com: Peter, the AI Pipeline Employee

Meet Peter, an AI employee who emails you a short list of hand-picked companies each morning and learns from every Keep, Skip, and Hold. The waitlist, a one-card-at-a-time pilot review, an endless interview called the River of Questions, and single sign-on are all live.

Visit it live ↗
PipelineRiver.com: Peter, the AI Pipeline Employee screenshot
First pilot batch reviewed at a 55% keep rate · 20-company curated queues · Peter reads and records every reply
⚡ An AI employee with his own inbox

The problem

Prospect lists are haystacks: thousands of rows, no judgment, and nobody actually works them. Joe wanted the opposite, a colleague who shows up every morning with a handful of companies worth a real look.

What Claude Code did

Claude built Peter end to end: a branded waitlist with priority scoring, welcome emails sent from Peter's own address, and inbound capture so replies to Peter land back in the system instead of vanishing. Pilot batches of 20 companies are reviewed one card at a time with a live preview of each prospect's website, and every verdict comes with a reason chip so Peter learns what to hunt for next. Batch two was curated entirely from batch one's verdicts. A 30-question River of Questions interview teaches Peter each user's business one wave at a time, and evyAI single sign-on guards the door.

The result

The waitlist went live the same evening the plan was finished, and the first signups arrived within hours, every one of them an existing customer. The first pilot batch came back with 11 keeps out of 20, and the second batch, built from those lessons, is in review.

Under the hood

ClaudeNode.js, zero dependenciesResend, inbound + outboundevyAI single sign-onApollo + Google Maps data

FAQ

Who exactly is Peter?

An AI employee with his own email address at pipelineriver.com. He sends the morning companies, and when you reply, he reads it and it goes on your record.

What do Keep and Skip actually do?

They are training data. Each verdict carries a reason chip, and the next batch is curated from what you kept and why. Batch two looked very different from batch one because of it.

Can I try it?

The waitlist is open at pipelineriver.com. Curated batches are rolling out to pilot users first.

Ask Joe about this project

Want to build things like this?Apply to the Claude Masterclass →